Use vec_extact_lo instead of subreg in reduc_<code>_scal_m.

Use vec_extact_lo instead of subreg in reduc_<code>_scal_m.

Loop vectorizer will use vec_perm to select lower part of a vector,
there could be some redundancy when using subreg in
reduc_<code>_scal_m, because rtl cse can't figure out vec_select lower
part is just subreg.

I'm trying to canonicalize vec_select to subreg like aarch64 did, but
there're so many regressions, some are easy to fix, some requires
middle-end adjustment.

So for simplicity, the patch use vec_select instead of subreg in
reduc_<code>_scal_m.

gcc/ChangeLog:

	* config/i386/sse.md: (reduc_plus_scal_<mode>): Use
	vec_extract_lo instead of subreg.
	(reduc_<code>_scal_<mode>): Ditto.
	(reduc_<code>_scal_<mode>): Ditto.
	(reduc_<code>_scal_<mode>): Ditto.
	(reduc_<code>_scal_<mode>): Ditto.